Conversation interaction method, system and equipment based on federal learning and storage medium

Abstract

The invention provides a dialogue interaction method, system and device based on federal learning and a storage medium. The dialogue interaction method comprises the steps that a user terminal sends a service request to a server side; the server side receives and reads user feature information in the service request, and transmits a feature portrait extraction model to the user terminal; in response to the blank user feature information, transmitting the general dialogue model to the user terminal; in response to non-blank user feature information, transmitting a personalized dialogue model corresponding to the user features; the user terminal receives the dialogue model and the feature portrait extraction model, the dialogue model is locally operated to serve a user, and the feature portrait extraction model is called to perform feature portrait extraction; and an update service request is sent to the server side again according to the extracted feature portrait, and personalized services are provided for the user by transmitting the personalized dialogue model. According to the method, personalized services are provided for different users on the premise that the privacy of the users is not violated, and the service expectation of the users is better met.

Description

Technical Field

[0001] This application belongs to the field of computer technology, and particularly relates to a dialogue interaction method, system, device, and storage medium based on federated learning. Background Art

[0002] Most current intelligent dialogue systems adopt a unified service model for all users. This model is usually trained based on a large-scale dataset and uses a unified algorithm and model to process and respond to user inputs. This unified service model can provide a wide range of functions and applications, but it cannot meet personalized needs and specific domain expertise, resulting in responses that may be too general or do not meet user expectations, leading to a decline in the service effect and efficiency for users.

[0003] In addition, users' backgrounds, interests, and needs also vary. For example, one user may be more interested in sports news, while another user may be interested in technology topics. Then, it is difficult for a unified model to provide personalized content in a specific domain for different users, thus unable to meet users' needs for in-depth communication and information acquisition in a specific domain. Additionally, personalized needs also include preferences for users' dialogue styles. Some users may prefer a humorous and relaxed dialogue style, while others may be more inclined to a formal and professional communication style. It is very difficult for a unified model to provide a personalized dialogue style according to different users' preferences, thus unable to meet users' personalized needs.

[0004] On the other hand, existing intelligent dialogue systems directly collect data such as user conversations for system training and iteration. During the iteration process of improving system performance, they cannot effectively meet users' privacy needs. Summary of the Invention

[0005] In view of this, this application aims to propose a dialogue interaction method, system, device, and storage medium based on federated learning to solve at least one of the above problems.

[0006] To achieve the above object, the technical solution of this application is realized as follows: In a first aspect, this application provides a dialogue interaction method based on federated learning, including: A user terminal sends a service request to a server terminal, where the service request contains blank user feature information and non-blank user feature information; The server terminal receives and reads the user feature information in the service request, and transmits a feature portrait extraction model to the user terminal; in response to blank user feature information, it transmits a general dialogue model to the user terminal; in response to non-blank user feature information, it transmits a personalized dialogue model corresponding to the user features; The user terminal receives the dialogue model and the feature portrait extraction model, locally runs the dialogue model to serve the user, and calls the feature portrait extraction model based on the local storage capacity to extract the feature portrait; The user terminal re - sends an updated service request to the server side according to the extracted feature portrait, and provides personalized services to the user by transmitting the personalized dialogue model; Based on the personalized service feedback result, update the dialogue model and the feature portrait extraction model to obtain model update parameters, and transmit them to the server side for update iteration after differential privacy randomization processing.

[0007] In a second aspect, based on the same inventive concept, the present application further provides a dialogue interaction device based on federated learning, including: A request sending module, configured to send a service request from a user terminal to a server side, where the service request includes blank user feature information and non - blank user feature information; A request receiving module, configured to receive and read the user feature information in the service request by the server side, and transmit the feature portrait extraction model to the user terminal; in response to blank user feature information, transmit the general dialogue model to the user terminal; in response to non - blank user feature information, transmit the personalized dialogue model corresponding to the user features; A portrait extraction module, configured to receive the dialogue model and the feature portrait extraction model by the user terminal, locally run the dialogue model to serve the user, and call the feature portrait extraction model based on the local storage capacity to extract the feature portrait; A personalized service module, configured to re - send an updated service request to the server side by the user terminal according to the extracted feature portrait, and provide personalized services to the user by transmitting the personalized dialogue model; A model update module, configured to update the dialogue model and the feature portrait extraction model based on the personalized service feedback result to obtain model update parameters, and transmit them to the server side for update iteration after differential privacy randomization processing.

[0008] In a third aspect, based on the same inventive concept, the present application further provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, where when the processor executes the program, the method described in the first aspect is implemented.

[0009] Fourthly, based on the same inventive concept, the present application also provides a non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores computer instructions for causing the computer to execute the method described in the first aspect.

[0010] Compared with the prior art, the dialogue interaction method, system, device and storage medium based on federated learning described in the present application have the following beneficial effects: For the dialogue interaction method, system, device and storage medium based on federated learning described in the present application, the method can provide personalized services for different users without infringing on user privacy, and better meet the service expectations of users. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The accompanying drawings, which form a part of this application, are used to provide a further understanding of the application. The illustrative embodiments and descriptions thereof of the application are used to explain the application and do not constitute an improper limitation of the application. In the drawings: Figure 1 is a flowchart of the dialogue interaction method based on federated learning according to an embodiment of the present application; Figure 2 is a schematic structural diagram of the dialogue interaction device based on federated learning according to an embodiment of the present application; Figure 3 is a schematic hardware structure diagram of the electronic device according to an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0012] In order to make the objectives, technical solutions and advantages of the present application more clearly understood, the following further describes the present application in detail with reference to specific embodiments and the accompanying drawings.

[0013] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the embodiments of the present application should have the ordinary meanings understood by those of ordinary skill in the art to which the present application belongs. The "first", "second" and similar terms used in the embodiments of the present application do not denote any order, quantity or importance, but are only used to distinguish different components. The terms such as "including" or "comprising" mean that the elements or objects appearing before this term cover the elements or objects listed after this term and their equivalents, without excluding other elements or objects. The terms such as "connected" or "coupled" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The terms such as "upper", "lower", "left" and "right" are only used to represent relative positional relationships, and when the absolute positions of the objects being described change, the relative positional relationships may also change accordingly.

[0014] Federated Learning moves the model training process to local devices, and only the updated parameters of the model are uploaded to the central server for aggregation to obtain the update of the global model. This not only solves the privacy and security issues but also enables personalized services for different users by distinguishing between the global model and the local model. In this embodiment, the general users are served by the aggregated global model, and when the users exhibit obvious category characteristics, personalized services are provided by training their local models.

[0015] The embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0016] Please refer to Figure 1 As shown, this embodiment provides a dialogue interaction method based on federated learning, which can provide personalized services for different users without infringing on user privacy, better meeting the service expectations of users. This method is applicable to various dialogue interaction system applications such as digital government affairs and online customer service, and specifically includes the following steps: Step S101: The user terminal sends a service request to the server terminal, where the service request contains blank user feature information and non-blank user feature information.

[0017] Step S102: The server terminal receives and reads the user feature information in the service request, and transmits the feature portrait extraction model to the user terminal; in response to blank user feature information, the general dialogue model is transmitted to the user terminal; in response to non-blank user feature information, the personalized dialogue model corresponding to the user features is transmitted.

[0018] Step S103: The user terminal receives the dialogue model and the feature portrait extraction model, locally runs the dialogue model to serve the user, and calls the feature portrait extraction model for feature portrait extraction based on the local storage capacity.

[0019] Step S104: The user terminal re-sends an updated service request to the server terminal according to the extracted feature portrait, and provides personalized services to the user by transmitting the personalized dialogue model.

[0020] Step S105: Update the dialogue model and the feature portrait extraction model based on the personalized service feedback result to obtain model update parameters, and transmit them to the server terminal for update iteration after differential privacy randomization processing.

[0021] This method obtains the user feature portrait from the user feature data. The update of each model adopts the method of federated learning. Through the model training framework of federated learning, it solves the problem that the existing methods collect training data and violate user privacy. The user only sends the model update parameters to the server side (the server side can provide both a generalized global model for user services and use a personalized model for user services according to the user portrait), without sending privacy data, effectively protecting the privacy of user data.

[0022] In addition, on the premise of the federated learning framework, differential privacy technology is adopted to protect the update parameters by randomization to prevent reverse user privacy information through the user's model update parameters, further improving the privacy protection performance.

[0023] In step S101 of this embodiment, a characteristic portrait of the user is constructed based on the user feature portrait model, that is, the personal characteristics of the user are modeled and described. This portrait can capture features such as the user's behavior pattern and interest preference, and perform federated learning training on the basis of protecting user privacy, with the user's personalized portrait as the output.

[0024] Through federated learning training, the user data of each participating party can be aggregated and analyzed on the premise of protecting user privacy for the training of this model. First, each participating party encrypts and anonymizes its own user data and transmits the encrypted data to the federated learning server; then, the federated learning server aggregates these data and uses the aggregated data for the modeling and training of the user characteristic portrait; finally, the trained user feature portrait model will be used for subsequent user feature portrait extraction.

[0025] At the same time, this embodiment provides an interaction interface between the user and the system, and locally stores the user's conversation data, which is input to the user feature portrait model running locally, so as to obtain the user's personalized user portrait and provide it to the server side for the server side to select the corresponding personalized service model to provide personalized services and recommendations for the user.

[0026] In step S102, a general dialogue model (global dialogue model) is established. This model is an initial dialogue model pre-trained based on a public dataset and provides general services as a basic model.

[0027] A personalized dialogue model is established. On the basis of the global model, combined with the user portrait and the user's conversation data, each user is individually fine-tuned to generate a personalized model.

[0028] The initial global dialogue model and personalized dialogue model are trained using public data sets and then put online for user services. During the user service process, all user session data obtains the updated parameters of the dialogue model through the user's local model update module and uploads it to the server. On the server side, these parameters are aggregated through federated learning technology to complete the online update iteration of the global dialogue model. In addition, the updated parameters of each user and their corresponding personalized portrait data will also be used for the update iteration of multiple personalized dialogue models.

[0029] The server generates a new dialogue model by aggregating model parameters (such as gradient or weight updates) uploaded by multiple users. The formula of the federated average algorithm is: In the formula, represents the updated model parameters, Indicates user The local model parameters, Indicates user The amount of local data, Indicates the total amount of data for all users.

[0030] Since the aggregated input is the model parameters of the user terminal rather than the original data, it breaks through the data silo problem and avoids privacy leakage.

[0031] In step S105, the parameter update data of the model is obtained locally using the user data, and the parameter update data is differentially protected and uploaded to the server for model update iteration. During the model update process, the differential privacy technology is used to randomize the update of the model parameters to protect the user's privacy information. Specific randomization methods may include adding noise, perturbing data, etc. Through differential privacy protection, malicious attackers can be prevented from obtaining the user's privacy information through the model update process, further improving the security of user privacy protection.

[0032] Embodiment 1: Based on the above method, take a single new blank user terminal access service as an example: 1) The user terminal initiates a service request to the server and includes blank user feature information in the request.

[0033] 2) The server receives the request from the user terminal, reads the blank user feature information in the request, and transmits the general dialogue model to the user terminal; at the same time, it transmits the feature portrait extraction model to the user terminal.

[0034] 3) The user terminal receives the general dialogue model and feature profile extraction model, and runs the general dialogue model locally to provide services to the user.

[0035] 4) The user terminal locally stores the conversation data and other user information (including privacy data) during the service process. When the data volume reaches a certain threshold, it calls the feature portrait extraction model to extract the user's feature portrait.

[0036] 5) Based on the extracted user feature portrait, the user terminal re-initiates an update service request to the server, including the extracted user feature information in the request.

[0037] 6) The server receives the update service request from the user terminal and transmits the corresponding personalized conversation model to the user terminal according to the user's feature information.

[0038] 7) The user terminal receives the personalized conversation model and calls the new personalized conversation model to serve the user.

[0039] 8) The user terminal locally stores the newly obtained conversation data and other user information during the service process. When the data update volume reaches the threshold, repeat steps 5)-7). At the same time, using the user's conversation data and user feedback, update the conversation model and the feature portrait extraction model to obtain model update parameters, and transmit them to the server after differential privacy randomization processing.

[0040] 9) The server trains and iterates its conversation model and feature portrait extraction model according to the received model update parameters, and updates the model parameters to obtain better model performance.

[0041] Example 2: Based on the above method, take the access of a user terminal with a single existing feature portrait to the service as an example: 1) The user terminal initiates a service request to the server, including the user feature information (i.e., including the existing feature portrait) in the request.

[0042] 2) The server receives the request from the user terminal, reads the user feature information in the request and transmits the personalized conversation model corresponding to the user feature; at the same time, transmits the feature portrait extraction model to the user terminal.

[0043] 3) The user terminal receives the personalized conversation model and the feature portrait extraction model, and locally runs the personalized conversation model to serve the user.

[0044] 4) The user terminal locally stores the conversation data and other user information during the service process. When the data volume reaches a certain threshold, it calls the feature portrait extraction model and combines the initial feature information to extract the user's feature portrait.

[0045] 5) Based on the updated user feature portrait, the user terminal re-initiates an update service request to the server, including the updated user feature information in the request.

[0046] 6) The server receives the update service request from the user terminal and transmits the updated personalized dialogue model to the user terminal according to the user's characteristic information.

[0047] 7) The user terminal receives the new personalized dialogue model and calls the new personalized dialogue model to serve the user.

[0048] 8) The user terminal locally stores the newly obtained dialogue data and other user information during the service process. When the data update volume reaches the threshold, steps 5)-7) are repeated; at the same time, using the user's dialogue data and user feedback, the dialogue model and the feature portrait extraction model are updated to obtain model update parameters, which are transmitted to the server side after differential privacy randomization processing.

[0049] 9) The server trains and iterates its dialogue model and feature portrait extraction model according to the received model update parameters, and updates the model parameters to obtain better model performance.

[0050] Embodiment 3: Based on the above method, taking the example of multiple user terminals accessing the service simultaneously: 1) Multiple user terminals initiate service requests to the server, and include user characteristic information in the requests.

[0051] 2) The server receives the requests from each user terminal, reads the user characteristic information in the requests. If it is blank, it transmits the general dialogue model to the corresponding user terminal; if it is not blank, it transmits the personalized dialogue model corresponding to the user characteristics; at the same time, it transmits the feature portrait extraction model to each user terminal. The feature portrait extraction model is transmitted to all users by broadcast, and the dialogue model is transmitted to each user by point-to-point communication.

[0052] 3) Each user terminal receives its own dialogue model and feature portrait extraction model, and locally runs the dialogue model to serve the user. Each user terminal locally stores the dialogue data and other user information during the service process. When the data volume reaches a certain threshold, it calls the feature portrait extraction model to update the user's feature portrait.

[0053] 5) The terminal with the updated user feature portrait re-initiates an update service request to the server, and includes the new user characteristic information in the request.

[0054] 6) The server receives the update service request from the user terminal and transmits the updated personalized dialogue model to the user terminal according to the user's characteristic information.

[0055] 7) The user terminal receives the personalized dialogue model and calls the new personalized dialogue model to serve the user.

[0056] 8) Each user terminal locally stores the newly acquired conversation data and other user information during the service process. When the data update amount reaches the threshold, steps 5)-7) are repeated. At the same time, using the user's conversation data and user feedback, the conversation model and the feature portrait extraction model are updated to obtain model update parameters, which are transmitted to the server side after differential privacy randomization processing.

[0057] 9) The server side trains and iterates its conversation model and feature portrait extraction model according to the received model update parameters, and updates the model parameters to obtain better model performance.

[0058] In addition, when the server side receives sufficient parameter updates from any user terminal, it actively initiates a model update request to all user terminals to transmit a new model to each user terminal.

[0059] It should be noted that some embodiments of the present application are described above. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the above embodiments and still achieve the desired results. Additionally, the processes depicted in the figures do not necessarily require the particular order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0060] Based on the same inventive concept, corresponding to the method of any of the above embodiments, an embodiment of the present application further provides a dialogue interaction device based on federated learning.

[0061] As Figure 2 shown, the dialogue interaction device based on federated learning includes: A request sending module 11, configured to send a service request from a user terminal to a server side, where the service request includes blank user feature information and non-blank user feature information; A request receiving module 12, configured to receive and read the user feature information in the service request by the server side, and transmit the feature portrait extraction model to the user terminal; in response to blank user feature information, transmit the general dialogue model to the user terminal; in response to non-blank user feature information, transmit the personalized dialogue model corresponding to the user feature; A portrait extraction module 13, configured to receive the dialogue model and the feature portrait extraction model by the user terminal, locally run the dialogue model to serve the user, and call the feature portrait extraction model to extract the feature portrait based on the local storage amount; A personalized service module 14, configured to re-send an updated service request to the server side by the user terminal according to the extracted feature portrait, and provide personalized service to the user by transmitting the personalized dialogue model; The model update module 15 is configured to update the dialogue model and the feature portrait extraction model based on the personalized service feedback result to obtain model update parameters, and transmit them to the server side for update iteration after differential privacy randomization processing.

[0062] For the convenience of description, when describing the above device, various modules are described separately according to their functions. Of course, when implementing the embodiments of the present application, the functions of each module can be implemented in the same or multiple software and / or hardware.

[0063] The device of the above embodiment is used to implement the corresponding method in any of the foregoing embodiments, and has the beneficial effects of the corresponding method embodiments, which will not be described in detail here.

[0064] Based on the same inventive concept, corresponding to the method of any of the above embodiments, an embodiment of the present application further provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, it implements the method described in any of the above embodiments.

[0065] Figure 3 FIG. shows a more specific schematic diagram of the hardware structure of the electronic device provided in this embodiment. The device may include: a processor 1010, a memory 1020, an input / output interface 1030, a communication interface 1040, and a bus 1050. Among them, the processor 1010, the memory 1020, the input / output interface 1030, and the communication interface 1040 are communicatively connected to each other inside the device through the bus 1050.

[0066] The processor 1010 may be implemented in a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, etc., and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0067] The memory 1020 may be implemented in the form of a ROM (Read Only Memory), a RAM (Random Access Memory), a static storage device, a dynamic storage device, etc. The memory 1020 may store an operating system and other application programs. When implementing the technical solutions provided in the embodiments of this specification through software or firmware, the relevant program codes are stored in the memory 1020 and are called and executed by the processor 1010.

[0068] The input / output interface 1030 is used to connect to the input / output module to achieve information input and output. The input / output module can be configured as a component in the device (not shown in the figure), or can be externally connected to the device to provide corresponding functions. Among them, the input device can include a keyboard, a mouse, a touch screen, a microphone, various sensors, etc., and the output device can include a display, a speaker, a vibrator, an indicator light, etc.

[0069] The communication interface 1040 is used to connect to the communication module (not shown in the figure) to achieve communication interaction between this device and other devices. Among them, the communication module can achieve communication through wired means (such as USB, network cable, etc.), or can achieve communication through wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0070] The bus 1050 includes a path for transmitting information between various components of the device (such as the processor 1010, the memory 1020, the input / output interface 1030, and the communication interface 1040).

[0071] It should be noted that although the above device only shows the processor 1010, the memory 1020, the input / output interface 1030, the communication interface 1040, and the bus 1050, in the specific implementation process, the device may also include other components necessary for normal operation. In addition, those skilled in the art can understand that the above device may also only include the components necessary to implement the solution of the embodiments of this specification, and do not have to include all the components shown in the figure.

[0072] The electronic device of the above embodiment is used to implement the corresponding method in any of the foregoing embodiments, and has the beneficial effects of the corresponding method embodiments, which will not be elaborated here.

[0073] Based on the same inventive concept, corresponding to the method in any of the above embodiments, the present application also provides a non-transitory computer-readable storage medium, and the non-transitory computer-readable storage medium stores computer instructions, and the computer instructions are used to make the computer execute the method described in any of the above embodiments.

[0074] The computer-readable medium of this embodiment includes both permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette tapes, magnetic tape disk storage or other magnetic storage devices, or any other non-transitory medium that can be used to store information that can be accessed by a computing device.

[0075] The computer instructions stored in the storage medium of the above embodiment are used to cause the computer to execute the method described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be elaborated here.

[0076] Those of ordinary skill in the art should understand that: the discussion of any of the above embodiments is only exemplary and is not intended to imply that the scope of the present application (including the claims) is limited to these examples; under the concept of the present application, the technical features in the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations in different aspects of the embodiments of the present application as described above, and they are not provided in detail for the sake of brevity.

[0077] In addition, for the sake of simplicity of description and discussion, and in order not to make the embodiments of the present application difficult to understand, the well-known power / ground connections to integrated circuit (IC) chips and other components may or may not be shown in the provided drawings. In addition, the devices may be shown in block diagram form in order to avoid making the embodiments of the present application difficult to understand, and this also takes into account the fact that the details of the implementation of these block diagram devices are highly dependent on the platform on which the embodiments of the present application will be implemented (i.e., these details should be fully within the understanding of those skilled in the art). In the case where specific details (such as circuits) are set forth to describe the exemplary embodiments of the present application, it will be apparent to those skilled in the art that the embodiments of the present application can be implemented without these specific details or with variations of these specific details. Therefore, these descriptions should be considered illustrative rather than restrictive.

[0078] Although the present application has been described in connection with specific embodiments thereof, many alternatives, modifications, and variations of these embodiments will be apparent to those of ordinary skill in the art in light of the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may be used with the embodiments discussed.

[0079] Embodiments of the present application are intended to cover all such alternatives, modifications, and variations that fall within the broad scope of the appended claims. Accordingly, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the embodiments of the present application shall be included within the protection scope of the present application.

Claims

1. A conversational interaction method based on federated learning, characterized in that: include: The user terminal sends a service request to the server, wherein the service request includes blank user characteristic information and non-blank user characteristic information; The server receives and reads the user feature information in the service request, and transmits a feature portrait extraction model to the user terminal; in response to blank user feature information, a general dialogue model is transmitted to the user terminal; in response to non-blank user feature information, a personalized dialogue model corresponding to the user feature is transmitted; The user terminal receives the conversation model and the feature portrait extraction model, locally runs the conversation model to provide services to the user, and calls the feature portrait extraction model to extract feature portraits based on the local storage capacity; The user terminal re-sends an update service request to the server according to the extracted feature portrait, and provides personalized services to the user by transmitting the personalized dialogue model; The dialogue model and the feature portrait extraction model are updated based on the personalized service feedback results to obtain model update parameters, which are then transmitted to the server for update and iteration after differential privacy randomization.

2. The method according to claim 1, characterized in that: The user characteristic information includes the user's conversation data and privacy data, and the user's personality portrait is obtained through an interactive interface based on the user's conversation data and privacy data.

3. The method according to claim 1, characterized in that: In response to a single user terminal and blank user feature information, a universal dialogue model is transmitted to the user terminal, the user terminal receives the universal dialogue model and the feature portrait extraction model, and locally runs the universal dialogue model to provide services to the user; The user terminal locally stores the conversation data and privacy data during the service process, and when the data volume reaches a certain threshold, calls the feature profile extraction model to extract feature profiles of the user.

4. The method according to claim 1, characterized in that: In response to a single user terminal and containing non-blank user feature information, a personalized dialogue model corresponding to the user feature is transmitted to the user terminal, the user terminal receives the personalized dialogue model and the feature portrait extraction model, and locally runs the personalized dialogue model to provide services to the user; The user terminal locally stores the conversation data and privacy data during the service process, and when the data volume reaches a certain threshold, calls the feature profile extraction model to extract feature profiles of the user.

5. The method according to claim 1, characterized in that: In response to the fact that there are multiple user terminals, the multiple user terminals send service requests to the server, the feature portrait extraction model is transmitted to all user terminals via broadcasting, and the dialogue model is transmitted to each user terminal concentrically via point-to-point transmission.

6. The method according to claim 1, characterized in that: The initial general dialogue model and personalized dialogue model are trained using public data sets and then put online for user service. During the user service process, all user session data obtains the updated parameters of the dialogue model through the user's local model update module and uploads it to the server. On the server side, the updated parameters are aggregated through the federated learning algorithm to perform online update and iteration of the general dialogue model and the personalized dialogue model.

7. The method according to claim 1, characterized in that The user terminal resends an update service request to the server according to the extracted feature portrait, and provides personalized services to the user by transmitting the personalized dialogue model, including: The user terminal re-sends an update service request to the server according to the extracted user feature portrait, and the request includes the extracted user feature information; The server receives the update service request from the user terminal and transmits the updated personalized dialogue model to the user terminal; The user terminal calls the new personalized dialogue model to provide services to the user.

8. A conversational interaction device based on federated learning, characterized in that: include: A request sending module, configured to send a service request from a user terminal to a server, wherein the service request includes blank user characteristic information and non-blank user characteristic information; The request receiving module is configured to receive and read the user feature information in the service request at the server side, and transmit the feature portrait extraction model to the user terminal; in response to blank user feature information, transmit the general dialogue model to the user terminal; in response to non-blank user feature information, transmit the personalized dialogue model corresponding to the user feature; A portrait extraction module is configured such that when the user terminal receives the conversation model and the feature portrait extraction model, the conversation model is locally run to provide services to the user, and the feature portrait extraction model is called based on the local storage amount to perform feature portrait extraction; A personalized service module, configured so that the user terminal resends an update service request to the server according to the extracted feature portrait, and provides personalized services to the user by transmitting the personalized dialogue model; The model update module is configured to update the dialogue model and the feature portrait extraction model based on the personalized service feedback results to obtain model update parameters, which are then transmitted to the server side for update and iteration after differential privacy randomization.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the method according to any one of claims 1 to 7 is implemented.

10. A non-transitory computer-readable storage medium, characterized in that: in, The non-transitory computer-readable storage medium stores computer instructions, and the computer instructions are used to cause a computer to execute the method according to any one of claims 1 to 7.

Images